klibc - Feb 2021 - [PATCH 0/1] workaround for overlapping sections in binary

Problem: binary (typesize.bin) created from object file (typesize.o)
with 'objcopy -O binary ... '. But typesize.o has relocatable objects
that all
copied with offset 0. This will lead to overlapping sections in binary.
By fortunate syscalls.pl check magic bytes and ring the bell.

Here suggested naive workaround: just skip .note.gnu.property section that
overlap
.rodata section (at least with objcopy from 2.35 and 2.36 binutils).

Guys from binutils team think that this not a objcopy problem:
https://sourceware.org/bugzilla/show_bug.cgi?id=27314

Related commit:

  commit de6f630e6be90d6d32d8bf2fed3f856b0c32f7ba
  Author: H. Peter Anvin <hpa at zytor.com>
  Date:   Sat Jun 10 11:15:19 2006 -0700

      [klibc] Detect the sizes of various types, and make available to
sysstub.ph.

      This additional code effectively queries the C compiler for the sizes of
      various types, and makes an associative array %typesize available to
      sysstub.ph.  This is currently not used, but it's expected that some
      architectures, e.g. s390, will need this to determine which registers
      go where, and how many registers are needed.


Petr Ovtchenkov (1):
  workaround for overlapping sections in binary

 usr/klibc/syscalls/Kbuild | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

-- 
2.29.2.1.g08c1adf7b9

Problem: binary (typesize.bin) created from object file (typesize.o)
with 'objcopy -O binary ... '. But typesize.o has relocatable objects
that all
copied with offset 0. This will lead to overlapping sections in binary.
By fortunate syscalls.pl check magic bytes and ring the bell.

This is naive workaround: skip .note.gnu.property section that overlap
.rodata section.

This not a bug of objcopy, https://sourceware.org/bugzilla/show_bug.cgi?id=27314

Related commit:

  commit de6f630e6be90d6d32d8bf2fed3f856b0c32f7ba
  Author: H. Peter Anvin <hpa at zytor.com>
  Date:   Sat Jun 10 11:15:19 2006 -0700

      [klibc] Detect the sizes of various types, and make available to
sysstub.ph.

      This additional code effectively queries the C compiler for the sizes of
      various types, and makes an associative array %typesize available to
      sysstub.ph.  This is currently not used, but it's expected that some
      architectures, e.g. s390, will need this to determine which registers
      go where, and how many registers are needed.
---
 usr/klibc/syscalls/Kbuild | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/usr/klibc/syscalls/Kbuild b/usr/klibc/syscalls/Kbuild
index 2430b9b4..754d028e 100644
--- a/usr/klibc/syscalls/Kbuild
+++ b/usr/klibc/syscalls/Kbuild
@@ -71,7 +71,7 @@ $(obj)/typesize.c: $(srctree)/$(KLIBCSRC)/syscalls.pl
$(obj)/SYSCALLS.i     \
 
 # Convert typesize.o to typesize.bin
 quiet_cmd_mkbin = OBJCOPY $@
-      cmd_mkbin = $(KLIBCOBJCOPY) -O binary $< $@
+      cmd_mkbin = $(KLIBCOBJCOPY) -O binary --remove-section .note.gnu.property
$< $@
 
 $(obj)/typesize.bin: $(obj)/typesize.o FORCE
 	$(call if_changed,mkbin)
-- 
2.29.2.1.g08c1adf7b9